US 20070044002 A1
An offload system, method, and computer program product are provided for utilizing a hardware network interface for identifying data and calculating at least a portion of a cyclical redundancy check (CRC) value for the data.
1. A method, comprising:
identifying data; and
calculating at least a portion of a cyclical redundancy check (CRC) value for the data, utilizing a hardware network interface.
2. The method of
3. The method of
4. The method of
5. The method of
6. The method of
7. A computer program product embodied on a tangible computer readable medium, comprising:
computer code that identifies data; and
computer code that calculates at least a portion of a cyclical redundancy check (CRC) value for the data, utilizing a hardware network interface.
8. The computer program product of
9. The computer program product of
10. The computer program product of
11. The computer program product of
12. The computer program product of
13. A system, comprising:
a hardware network interface operable to identify data;
the hardware network interface further operable to calculate at least a portion of a cyclical redundancy check (CRC) value for the data.
14. The system of
15. The system of
16. The system of
17. The system of
18. The system of
The present application is a continuation of U.S. application Ser. No. 10/851,357, filed May 20, 2004, which is a continuation of U.S. application Ser. No. 09/321,902, filed May 28, 1999 (now issued under U.S. Pat. No. 6,765,901), which claims priority from U.S. provisional application 60/088,865, filed Jun. 11, 1998, which are all incorporated herein by reference.
The invention relates to combining a network stack with a modem core for use in both computer and non-computer applications. More particularly, the invention relates to an Internet-aware modem which combines any number of point-to-point devices with the network protocols necessary to communicate on the Internet, where these devices include various speed traditional modems from 2400 kbps to 56 kbps, ISDN modems, newer xDSL modems, and digital cellular modems.
Computer modems were developed in a time when most connections were made to proprietary online services, interactive terminals, bulletin board services (BBSs), or corporate network systems. As such, it was necessary to implement connection protocols in software because there existed at the time a number of such protocols. These protocols included x-modem, y-modem, z-modem, kermit, and interactive character based interfaces.
Today, with popularity of the Internet, a vast majority of modems are now used exclusively to connect with ISP's, which in turn connect the user to the Internet. Therefore, there is now a predominant set of connection protocols. Such protocols are used for most modem connections. Accordingly, there is a real need and advantage in designing a modem that is Internet-ready.
The connection protocols used by the Internet and their hierarchical relationship are shown in
Current computer systems treat a modem subsystem as a serial device. A block diagram of an existing system is shown in
For received data, the modem first places all incoming packets into the input FIFO 26. The device can then be configured to interrupt the host CPU when any data are available or when the received data reaches some level (i.e. 16 bytes). When notified, the CPU reads all the data in the input FIFO, and stores the data temporarily in a buffer in system memory (not shown). The bottom protocol, PPP (see
Once the whole packet is received, the PPP portion of the software network stack passes the data up to the second protocol (IP). The IP software then processes the IP header and, after verifying the header checksum, passes the packet to the TCP handler. The TCP handler then checks its checksum, and passes the data on to the appropriate application, as specified by the PORT number in the TCP header.
Because most modems in computers today are used to connect to the Internet, it makes it economically feasible and practical to optimize a modem for this environment. What this entails is embedding in the modem system, the ability to handle the necessary network protocols and use the knowledge of the protocols to tune the transmission characteristics of the modem. This is the same rationale behind the popularity of Window's accelerator graphics cards. Because graphic chip manufacturers know that a vast majority of PC's today run the Microsoft Windows® operating system, they fine-tune their architectures to enhance the performance in this environment. This would not be practical if there were a number of operating systems with different graphic APIs, each with a significant portion of the market place. However, with the one over-riding standard, most graphic card manufactures have chosen to optimize their hardware for the Windows environment, even though today's Pentium class processors are very capable of handling the graphic chores without external support. This is because the function is required in most circumstances, and it is advantageous to offload the host processor so that it has that much more MIPs to run standard applications.
A similar situation now exists in the modem card market. It would therefore be advantageous to embed the Internet network protocol stack, along with special logic, thereby enabling the modem device to become Internet-ready, such that the modem system offloads much of the network protocol processing from the main CPU, while improving the overall performance of the communication system.
The invention embeds the Internet network protocol stack, along with special logic, thereby enabling the modem device to become Internet-ready. As a result, the modem system offloads much of the network protocol processing from the main CPU and improves the overall performance of the communication system. The invention provides an Internet-aware modem which combines any number of point-to-point devices with the network protocols necessary to communicate on the Internet, where these devices include various speed traditional modems from 2400 kbps to 56 kbps, ISDN modems, newer xDSL modems, and digital cellular modems.
In a system equipped with an Internet-ready modem, the Internet application first sets up the socket parameters. These include the destination port number, the type of connection (TCP/UDP), the TOS (Type-Of-Service) requirements, and the destination IP address. When the network stack on an Internet-ready modem card gets this information, it attempts to start a connection by sending out a SYN packet. This packet is passed to the IP engine, which attaches the IP header and calculates the IP header checksum. The packet is then passed to the PPP handler which attaches the PPP header, appends the PPP checksum, and escapes the data. After PPP encapsulation, the resulting network packet is sent through the output FIFO to the modem core. For this packet, the TCP engine indicates to the packet analyzer block that it is a SYN packet. The packet analyzer then indicates to the modem that this is a stand alone packet and that it can be sent immediately. Upon receiving this information, the modem sends the network packet out without first waiting the normal 50 ms to see if additional information is forthcoming.
After the destination socket sends a return SYN-ACK packet, an ACK packet is sent from the modem card. This packet follows the same steps as those for the SYN packet. At this point, the socket connection is up, and the application software (such as a Web browser) is notified. The application can now send its data directly to the modem in a data packet format. In this example, where the application is a Web browser, the application can send an HTTP request directly to the modem system via a packet interface as opposed to the serial port I/O interface in a regular modem system. DMA style data transports can be used for this purpose. In this method, a data byte count is programmed into the packet interface. Data can then be automatically transferred from memory into a modem card without further intervention from the host CPU. After all the data have been transferred, an interrupt from the modem card can be sent to the host CPU indicating that the data transfer is complete.
As the data arrive at the modem card, they are sent (in this example) to a TCP data output buffer. After all the data are received or when the maximum data size per packet has been received, the TCP block begins calculating the checksum. The packet is encapsulated in the same way as the SYN packet. In parallel to this, the TCP engine indicates to the packet analyzer block that the destination port for the packet is 80, which is the well-known HTTP port. The packet analyzer then knows that there are no more data and, again, the modem should send the current packet immediately.
When receiving network packets, the data are sent from the modem core through the input FIFO to the PPP engine, which parses the PPP header, unescapes the data, and starts a running check on the packet for checksum calculations. If the engine determines that the encapsulated data is an IP packet, it enables the IP engine, and all data past the PPP header is forwarded to the IP engine. The IP engine parses the IP header, checks the checksum, and if it determines that the encapsulated protocol is TCP, then it sends all data past the IP header to the TCP engine. The TCP engine then parses the TCP header and sends the data portion to the security layer. If the data is HTML data, it can be passed through a ratings check that parses out rating tags of pages. If the page has a rating below or equal to the modem cards setting, then the data are allowed to pass. If the rating exceeds the setting on the card, a message indicating so is passed on instead. If the page contains no ratings, a bit can be set to either pass or block the page. All non-HTML data are passed directly to the TCP data buffer.
As the data are being written into the buffer, a running count is kept to see how much data have been received. At the end of the network packet, if the PPP checksum indicates that the entire packet was received without errors, then an interrupt can be generated to the host CPU. The application can then read the received data count, and program a DMA transfer to transfer data from the TCP buffer into main memory.
Another Embodiment of the Invention
The invention provides a low-cost, low-power, easily manufacturable, small form-factor network access module which has a low memory demand and provides a highly efficient protocol decode. The invention comprises a hardware-integrated system that both decodes multiple network protocols in a byte-streaming manner concurrently and processes packet data in one pass, thereby reducing system memory and form factor requirements, while also eliminating software CPU overhead.
The preferred form of this embodiment of the invention comprises a network protocol layer, data handler, O.S. State Machine, and memory manager state machines implemented at a hardware gate level. Network packets are received from a physical transport level mechanism by the network protocol layer state machine. The protocol state machine decodes network protocols such as TCP, IP, User Datagram Protocol (UDP), PPP, and Raw Socket concurrently as each byte is received. Each protocol handler parses, interprets, and strips header information immediately from the packet, requiring no intermediate memory. The resulting data are passed to the next protocol layer or data handler for which the latter case consists of data state machines that decode data formats such as email, graphics, Hypertext Transfer Protocol (HTTP), Java, and Hypertext Markup Language (HTML). Each data state machine reacts accordingly to the pertinent data, and any data that are required by more than one data state machine are provided to each state machine concurrently. Any data that are required more than once by a specific data state machine, are placed in a specific memory location with a pointer designating such data (thereby ensuring minimal memory usage). Resulting display data are immediately passed preformatted to a display controller. Any outgoing network packets are created by the data state machines and passed through the network protocol state machine which adds formats to the packet, and checksums the information header information, and forwards the resulting network packet via a physical transport level mechanism.
The preferred form of this embodiment does not necessarily require a CPU and software to process the network packets, thereby greatly reducing system cost. The hardware gate level implementation provides a modular, embeddable design whereupon the designer may pick and choose the functionality that the particular application requires and still retain a low cost, low power, small form factor.
The invention provides an Internet-aware modem which combines any number of point-to-point devices with the network protocols necessary to communicate on the Internet, where these devices include various speed traditional modems from 2400 kbps to 56 kbps, ISDN modems, newer xDSL modems, and digital cellular modems. The invention embeds the Internet network protocol stack, along with special logic, thereby enabling the modem device to become Internet-ready. As a result, the modem system offloads much of the network protocol processing from the main CPU and improves the overall performance of the communication subsystem.
The invention provides several technologies that enhance modem performance and efficiency when used with Internet protocols. These technologies include:
Block Based Communication,
Reduced Processing Power,
Reduced Energy Requirements, and
Data Transfer Overhead Reduction
Modems today all communicate via serial ports. This serial communication has several performance disadvantages over other communications devices, such as Ethernet adapters, which communicate to the host CPU by sending and receiving blocks of data via DMA (Direct Memory Access).
Serial communication using older serial hardware (e.g. UARTs) require an interrupt for every character sent or received. This causes so much overhead that it is not possible to communicate at speeds over 19200 bps without dropping data on most computer systems. Second generation serial hardware (e.g. the 16550 UART and its derivatives) are able to buffer up to 16 bytes on send and receive and can delay interrupts until a buffer reaches a certain level. This reduces the number of interrupts required to transfer data to and from the modem. This allows modern modems to achieve serial data rates of 56000 bps to 230400 bps without losing data on most CPU's.
Although high speed serial communication using second generation serial hardware is now possible, at these higher data rates serial communications can cause a noticeable degradation in system performance, especially when the computer system is running Internet enabled action games or multi-media communication programs. Thousands of interrupts each second are required at the higher data rates (see Table 1 below.)
Each interrupt requires an interrupt service routine to read and write data to/from the serial hardware and to/from the Computers Operating System. These interrupt routines read and write to I/O ports that are not efficient use of system resources.
Optimizing Using Block Transfers
There has been talk about a next generation UART that would have either 32 or 64 byte buffers to reduce the number of interrupts and load on a system. While this may be helpful, the inventor's have recognized that the correct solution is to optimize the modem for Internet protocols and transfer data as blocks of data packets.
Block based Communication to/from modems has not been previously considered because the Internet protocols designed to interface a modem/serial device are serial based (SLIP/PPP) and are traditionally implemented in the computer system's operating system. Using the herein described invention, it is advantageous to move the network stack protocols from the computer operating system into the modem. The layers of the network protocol that can be implemented in a hardware device in accordance with the invention can range from just the PPP layer all the way to the IP, TCP, and UDP layers.
Example: Block Based Interrupt Requirement
Assuming an average 100-byte IP packet:
As is shown in the above example, transferring blocks instead of chunks of characters, as in the serial solution, reduces the number of interrupts required. An added benefit is that the transfer routine can use direct memory access (DMA) to pass data instead of slower I/O port pumping. Currently serial port hardware and software do not support DMA.
This enables the development of software that can produce faster, more efficient data handling routines because the host CPU is relieved of handling transferring of individual bytes of data.
Once the modem knows that it is dealing with IP packets, it has knowledge of where data units start and stop. Modem protocols, such as V.42 and V.42bis, can be optimized to take advantage of this knowledge.
V.42 has an outdated view of what kind of data are being carried over a modem, but if the network stack is embedded alongside a modem core, it can indicate to the modem core the end of data units. The modem core can then segment its V.42 packets accordingly. This reduces latency on retransmission because retransmission would have a higher probability of affecting only one IP packet. Knowledge of the end of a packet can also be used to reduce the waiting for more data timeout that the V.42 protocol introduces. If it is known that the last byte received is an end of a physical block of data, the modem need not wait around, or can reduce the time it waits, for more data to be sent to it before it sends that data it has ready to send.
This aspect of the invention can be extended to perform IP and TCP packet snooping to optimize further what gets sent when and how long the V.42 protocol waits before compressing and sending the current block. For example, the TOS field in the IP header can be used to determine the amount of latency used in the transmission of the packet. If the packet is a high priority packet the system may decide to send the packet immediately, not checking to determine if there are more packets ready to send. The system can wait varying lengths of time for more data based on this TOS information.
Additional latency optimization can be achieved by checking the TCP header. If the SYN flag is set in the TCP header, then the data should also be sent immediately because nothing more can happen until the SYN-ACK is returned from the other side of the connection.
Also, unlike previous advances to modem technology that required that the same technical advances be on both the receiving and sending ends of the communications link, the herein disclosed modem can operate and is compatible with all existing modems today. Therefore, it is possible to gain performance increases by updating only one end of the connection. In this way, adoption of the technology is not dependent on changes in ISP infrastructure.
Reduced Processing Power Required from Host CPU
By embedding the network stack along with the modem core, a marked reduction in processing power that is required to connect and communicate on the internet is possible. This allows small, low power, low cost devices to communicate via the Internet using modems. Examples of these types of devices include game consoles, PDAs, toys, and other consumer electronic and mobile electronic appliances.
Processor Power Reduction of PPP
The PPP protocol requires that packets include a CRC appended at the end of each packet. This calculation requires what could be a significant amount of processing power on low-end processors. Other aspects of the protocol, such as escaping data and parameter negotiations, require memory accesses for each byte when implemented as a software solution. By implementing PPP in the modem, all negotiations can be kept local to the modem subsystem, thereby relieving system bus traffic and processing overhead.
Processor Power Reduction of Embedding IP
Embedding IP offloads the header checksum calculations from the host processor. It also keeps ICMP echo packet processing local to the modem card. This protocol is used for PING applications.
Processor Power Reduction of Embedding UDP
Though it is possible to use UDP as an encapsulation device requiring little processing power, when used with checksums, UDP can require some processor resources. It should be noted here that most thin Internet clients are based on IP/UDP to transmit their data.
Processor Power Reduction of Embedding TCP
TCP is a much more complicated protocol than UDP and, thus, requires much more processing power. The TCP has many states and requires checksums to be preformed on packets. For embedded products that require TCP support the invention described herein provides a way to offload all the complexity and processor requirements onto a dedicated hardware circuit.
Portability of Solution
By embedding the network stack inside of the modem subsystem, the same modem system can now be used across multiple computing and system platforms. Because no porting of any network stack software is required, moving the modem into different systems becomes very easy. This is especially important in the embedded systems market that does not have one or two main OS's but instead is made up of a number of different OS's, RTOS's, and in some cases, no OS. The embedded systems market is also characterized by being made up of a number of incompatible processors. This lack of an overriding standard also favors a highly portable network solution.
Reduced Energy Requirements
The herein described invention is also very efficient in terms of energy requirements. A highly optimized state machine reduces by two orders of magnitude the clock rate required to perform the functionality of the Internet suite of protocols. This translates into extended battery life and less heat generated by products designed with the herein described invention.
By implementing the network stack in hardware along with the modem, the invention provides a very secure, unhackable network stack. This is due to the hardware architecture implementation that disregards any packet received unless there is already a socket connection set up for it. Furthermore, the packets do not get past the modem card, therefore making any interaction between unrequested packets and software impossible.
In addition, by including an HTML packet sniffer, it is possible to decode HTML rating tags. The packet sniffer interprets bytes in the packet buffer, and can be set up to pass only those pages that are within its preset rating level. For pages that exceed the programmed rating level, the HTML sniffer passes up a failed retrieval message, and does not allow the HTML content past the modem subsystem. This turns the modem into a content driven mini-firewall.
For those pages that do not include a rating tag, the sniffer can be configured either not to pass these pages at all, or to allow the pages to be passed. The rating level can be programmed only via board settings, i.e. hard wired. The advantage of implementing this solution in hardware is that for parents or anyone who wishes to filter out certain Web sites, it is impossible to get around this system without taking the modem card out of the system. With any software solution, the user could simply load a non-filtering browser or disable a certain plug-in, and the filter would be bypassed. By providing the filter in hardware on the modem card, there is no way to bypass the function.
In a system equipped with an Internet-ready modem 31, the Internet application first sets up the socket parameters. These include the destination port number, the type of connection (TCP/UDP), the TOS (Type-Of-Service) requirements, and the destination IP address. When the network stack 30 on the Internet-ready modem 31 gets this information, it attempts to start a connection by sending out a SYN packet. This packet is passed to the IP engine within the network stack, which attaches the IP header and calculates the IP header checksum. The packet is then passed to the PPP handler within the network stack which attaches the PPP header, appends the PPP checksum, and escapes the data. After PPP encapsulation, the resulting network packet is sent through the output FIFO 24 to the modem core 25. For this packet, the TCP engine indicates to the packet analyzer block 34 that it is a SYN packet. The packet analyzer 34 then indicates to the modem that this is a stand alone packet and that it can be sent immediately. The modem upon receiving this information sends the network packet out without waiting the normal 50 ms to see if additional information is forthcoming.
After the destination socket sends a return SYN-ACK packet, an ACK packet is sent from the modem card. This packet follows the same steps as those for the SYN packet. At this point, the socket connection is up, and the application software (such as a Web browser 21) is notified. The application can now send its data directly to the modem in a data packet format.
In the example shown in
As the data arrive at the modem card, they are sent (in this example) to a TCP data output buffer. After all the data are received or when the maximum data size per packet has been received, the TCP block begins calculating the checksum. The packet is encapsulated in the same way as the SYN packet. In parallel to this, the TCP engine in the network stack indicates to the packet analyzer block that the destination port for the packet is 80, which is the well-known HTTP port. The packet analyzer then knows that there are no more data and, again, the modem should send the current packet immediately.
When receiving network packets, the data are sent from the modem core 25 through the input FIFO 26 to the PPP engine in the network stack, which parses the PPP header, unescapes the data, and starts a running check on the packet for checksum calculations. If the engine determines that the encapsulated data is an IP packet, it enables the IP engine in the network stack, and all data past the PPP header is forwarded to the IP engine. The IP engine parses the IP header, checks the checksum, and if it determines that the encapsulated protocol is TCP, then it sends all data past the IP header to the TCP engine in the network stack. The TCP engine then parses the TCP header and sends the data portion to the security layer 36. If the data is HTML data, it can be passed through a ratings check that parses out rating tags of pages. If the page has a rating below or equal to the modem cards setting, then the data are allowed to pass. If the rating exceeds the setting on the card, a message indicating so is passed on instead. If the page contains no ratings, a bit can be set to either pass or block the page. All non-HTML data are passed directly to the TCP data buffer 37.
As the data are being written into the buffer, a running count is kept to see how much data have been received. At the end of the network packet, if the PPP checksum indicates that the entire packet was received without errors, then an interrupt can be generated to the host CPU 22. The application can then read the received data count, and program a DMA transfer to transfer data from the TCP buffer into main memory.
Features of the Invention
The following discussion describes various features of the invention:
1. Modem as a Block Device.
2. Latency Optimization based on packet parameters:
a) Basic End of Packet Optimization,
b) Optimization based on IP TOS flag,
c) Optimization based on UDP/TCP port numbers,
d) Optimization based on TCP State, and
e) Latency Table.
3. Modem as a complete Internet Access Device:
a) Partial Stack solutions:
b) Complete Stack Solutions (PPP/IP/ICMP/UDP/TCP).
4. Enhanced Security and HTML filtering in an Internet enabled Modem.
The following discussion describes the above features in more detail.
Modem as a Block Device
Depending on the network layers included with the modem hardware, different data formats are sent to the modem subsystem. In any of the implementations, however, DMA transfers can be used to optimize CPU overhead required for the transfers.
At a minimum, just PPP can be added to the modem. The function of PPP is to transform IP packets (blocks of data) into a serial stream so that it can be transported over a serial device (see
As with an Ethernet card, packets are shipped from the IP protocol stack to the device driver. In the modem's case, the device driver is a simple interfacing software program that transfers blocks of data to and from the modem and the host computer. Compare the architecture of the prior art modem of
This embodiment of the invention enables all of the efficiency benefits described above, and is an attractive solution because it can be implemented by adding only a minimum amount of extra logic to existing modem chip sets, and because it requires as little as 512 bytes of memory for support. This makes it very cost affordable to add to any existing modem.
Traditional modems have no knowledge of the type of data they carry and their protocols are optimized for interactive character based interfaces. The traditional modem protocols have a built in 50 ms delay before sending information. This delay is present because the modem has no idea of where the data stops and starts, so it waits until it knows no more data are going to be sent.
With the popularity of the Internet, almost all modems today are used to connect to the Internet. Using this knowledge and the information on when the packets begin and end can help optimize modem transmissions. This optimization can reduce the amount of time it takes to move an Internet protocol packet from one modem to another by reducing or eliminating the 50 ms delay built into the modem protocols. This feature of the invention can be extended using different parts of the network packet to make decisions whether there should be any delay or how long it should be before the modem processes a packet. The hardware module that handles this optimization is the packet analyzer block 34 (see
Basic End of Packet Optimization
At the most basic level, one could use the knowledge of the end of the PPP packet to tell the modem protocols to wait a small amount of time before sending the packet (see
This algorithm is useful for PPP sub-protocols, such as LCP (Link Control Protocol), PAP (password Authentication Protocol), CHAP (Challenge Handshake Authentication Protocol), and NCP (Network Control Protocol). With these and similar PPP sub-protocols, packets transmitted are stand-alone in that all information is contained within a single packet. Also, in most cases after the packet is sent, no further data are sent because the device is waiting for a response from the other device. Therefore, if the packet analyzer detects that a PPP packet contains a PPP sub-protocol, when it detects the PPP FCS field it can instruct the modem to wait only 2 ms before sending the data, instead of the normal 50 ms. The reason to wait a minimum of at least 2 ms is that in the transition between the LCP, Authentication, and NCP phases of the PPP negotiations, back to back packets can be sent out. However, there would never be more than two back to back packets, and the second packet always follows immediately within 2 ms of the first packet.
Further optimization can occur by looking at the command code of the PPP sub-protocol packet. An example matrix of command types and the corresponding latency setting are shown in the Table 2 below.
Optimizing Based on the TOS Field
FOR IP, TCP, and UDP packets, a more intelligent decision on how long to wait or when to send packets can be determined by examining the type of service (TOS) field in the IP header. The TOS field describes the priority and reliability requested for the packet. The properties that are settable for the TOS field are Minimize Delay, Maximize Throughput, Maximize Reliability, and Minimize Cost. More than one of the TOS flag properties can be set at one time. This information can be used to set variable waiting time delays or to send what is in the modem buffer immediately.
Optimizing Based on the Encapsulated Protocol
Another optimization that can be performed based on IP header fields is to base the latency setting on the protocol field. Most ICMP and IGMP packets are self contained therefore minimum wait times are needed after they are sent. After the packet analyzer determines that the IP packet contains either an IGMP or ICMP packet, it signals to the modem core to send the packet immediately.
Optimization Based on Packet Ports
Certain kinds of Internet services have information distributions that require just one packet of data to be sent and received. With these types of services it is optimal always to send a packet immediately without waiting for more data. Other Internet services have yet other packet distribution patterns that could be optimized for UDP and TCP, the major protocols that are used on top of IP. Both use ports to describe services.
The latency table 90 contains a table of ports and the amount of time to wait after the end of the packet for more data. An example of optimizing using this method is the DNS application. In this application, the entire data portion of the message can easily fit in one Internet packet. Therefore, if after examining the destination port in the UDP header it is determined that it is a DNS packet, the packet can be sent immediately because there are not any further packets coming. Table 3 below provides an example protocol-port latency table.
Latency Optimization Based on Packet State
TCP is a state-based protocol and certain states have well known properties that can be used for latency optimization. An example is the three-way handshake with which all TCP connections start. The first few packets of this transaction are small packets that must be sent before any further communications can take place. The invention can noticeably improve the time this connection process takes. This can be a very noticeable improvement, especially when operating software that connects many TCP connections at one time during one transaction, as with a Web browser.
Latency tables are state machines that have a number of inputs that are triggered by a packet's characteristics. From these inputs the latency table produces an optimized latency value for the modem protocol, effectively optimizing each packet as it passes through the system. A block schematic diagram of the latency tables is shown in the
The IP latency resolver 110 takes the inputs from the IP sub-protocol latency table 111 and the IP TOS field latency table 112 and selects the lower of the two values. The TCP latency resolver 113 performs a similar function for the destination port latency table 114 and the TCP state latency table 115. The IP latency resolver and TCP latency resolver outputs are muxed 117 to produce a combined latency value therefor. (The mux selection is determined by the protocol field in the IP header as parsed by the IP Sub-Protocol Latency Table 111.) A value is also provided by the PPP latency table 116. This value is muxed with the muxed 117 value of the IP latency resolver and TCP latency resolver. The mux selection for mux 118 is determined by the protocol field in the PPP header as parsed by the PPP latency table 116. The final latency value is then sent to the modem subsystem.
Modem as a Complete Internet Access Device
The extended Internet modem model is a stand alone, embeddable, highly integrated communications product that can Internet enable almost any electronic device (see
Enhanced Security Benefits
As stated above, one security benefit of having a hardware based network stack is that only those packets received that are destined for a preconfigured socket connection are allowed past the modem subsystem. All other packets are filtered out at the hardware level, making any interaction between these packets and software impossible. Also, with the addition of the HTML sniffer, V-Chip like filtering can be provided that cannot be easily circumvented. Block diagrams of the HTML sniffer are shown in
As shown in
Although the invention is described herein with reference to the preferred embodiment, one skilled in the art will readily appreciate that other applications may be substituted for those set forth herein without departing from the spirit and scope of the present invention. Accordingly, the invention should only be limited by the Claims included below.